thompson - brain in a vat or body in a world

PHILOSOPHICAL TOPICS

VOL. 39, NO. 1, SPRING 2011

Brain in a Vat or Body in a World?Brainbound versus Enactive Views of Experience

Evan ThompsonUniversity of Toronto

Diego CosmelliPontificia Universidad Católica de Chile

ABSTRACT. We argue that the minimal biological requirements for con-sciousness include a living body, not just neuronal processes in the skull.Our argument proceeds by reconsidering the brain-in-a-vat thoughtexperiment. Careful examination of this thought experiment indicates thatthe null hypothesis is that any adequately functional “vat” would be a sur-rogate body, that is, that the so-called vat would be no vat at all, but ratheran embodied agent in the world. Thus, what the thought experiment actu-ally shows is that the brain and body are so deeply entangled, structurallyand dynamically, that they are explanatorily inseparable. Such entangle-ment implies that we cannot understand consciousness by consideringonly the activity of neurons apart from the body, and hence we have goodexplanatory grounds for supposing that the minimal realizing system forconsciousness includes the body and not just the brain. In this way, we putthe brain-in-a-vat thought experiment to a new use, one that supports the“enactive” view that consciousness is a life-regulation process of the wholeorganism interacting with its environment.

163

Topics 39.1 first pages:Layout 1 3/25/13 4:58 PM 63

Is consciousness all in the head, or more precisely all in the brain? Or is the bodybeyond the brain an essential part of the biological basis of consciousness? To putthe question another way, does the body belong to the “minimal realizing system”for conscious experience or is the minimal realizing system for consciousness con-fined to the brain?

According to the “enactive” view of experience, consciousness is a life-regula-tion process of the body interacting with its environment.1 Perception, action, emo-tion, imagination, memory, dreaming—these are modes of self-regulation thatdepend directly on the living body and not just the brain. According to the enactiveview, the body shouldn’t be seen as a mere outside causal influence on an exclu-sively neuronal system for consciousness because the minimal requirements forconsciousness include a living body, not just neuronal events in the skull.

The enactive view stands in tension with the widespread view in the neuro-science of consciousness that consciousness is brainbound. Many neuroscientistsand philosophers would say that your brain directly determines what you experi-ence, but your body affects what you experience only via its influence on yourbrain. According to this way of thinking, the body is strictly inessential for con-scious experience; for example, in principle, a disembodied brain in a vat couldhave the same kinds of subjective experiences or states of phenomenal conscious-ness as an embodied brain.

Our focus will be on the tension between these two views—between what we’llcall the enactive view and the brainbound view, or Enactive and Brainbound forshort.2 As we’ll see, these views work with different conceptions of consciousness,and these different conceptions of consciousness motivate different workingassumptions about the biological basis of consciousness. Our aim is to persuadeyou to prefer Enactive to Brainbound. Specifically, we’ll argue that a strong case canbe made for thinking that the biological system for consciousness isn’t limited tothe brain. Consciousness isn’t a strictly brain phenomenon but an organism phe-nomenon. The minimal biological basis for consciousness includes nonneural fac-tors and physiological processes beyond the skull.

I. BRAINBOUND BASICS

We begin with Ned Block’s statement of what he calls the “orthodox view” of thebrain basis of consciousness, the view we’re calling Brainbound.3

Block presents Brainbound in the context of criticizing Alva Noë’s version ofthe enactive view.4 According to Noë, perceptual experience is a mode of tempo-rally extended skillful interaction with the world. To perceive is to explore one’senvironment by exercising in a practical and bodily way one’s sensorimotor knowl-edge of how sensory appearances vary as a function of movement. It follows, forNoë, that, “To perceive like us . . . you must have a body like ours.” As he says, “If

164

Topics 39.1 first pages:Layout 1 3/25/13 4:58 PM Page 164

perception is in part constituted by our possession and exercise of bodily skills . . .then it may also depend on our possession of the sorts of bodies that can encom-pass those skills, for only a creature with such a body could have those skills.”5

Block contends that this view misidentifies what causes experience with whatconstitutes experience. Although perceptual experience depends causally on havinga body and exercising sensorimotor know-how in movement, exercising this bodilyknowledge doesn’t constitute or directly determine perceptual experience. Rather,what directly determines a given perceptual experience is a specific pattern of brainactivity. If the right neural pattern were to occur, so too would the subjective expe-rience, regardless of the wider context of bodily activity in the environment.

Block interprets Noë as making a metaphysical claim about the subpersonal orphysical basis of perceptual experience—that it includes the body and not just thebrain. This interpretation strikes us as off-key. Noë’s enactive view and enactiveviews in general offer an explanatory framework for perception. We use phenome-nological considerations about perceptual experience to constrain how we thinkabout the subpersonal mechanisms of perception. First, we start by trying to get thephenomenology right, or to put it another way, by trying to conceptualize percep-tual experience properly at the personal level. Here the central idea is that to per-ceive is to be in an interactive relationship with the world, not to be in an internalstate that happens to be caused by the external world. Second, we argue that, giventhis conception of perceptual experience, we can’t specify the mechanisms of per-ception only in terms of what goes on in the brain without including the body andits dynamic sensorimotor coupling with the environment. Therefore, it makes littlesense to restrict the physical basis of perceptual experience to the brain withoutincluding the body.

Let’s go back to Brainbound. Block reads the enactive view as making a meta-physical claim about what he calls the “minimal constitutive supervenience base”for experience. In his words:

The issue of the constitutive supervenience base for experience is theissue of what is—and is not—a metaphysically necessary part of a meta-physically sufficient condition of perceptual experience. That is, it is theissue of what is—and is not—part of the minimal metaphysically suffi-cient condition for perceptual experience (the minimal superveniencebase). Noë’s enactive view says that the skilled active body is part of thatminimal condition (minimal supervenience base), whereas the viewwhich I hold and which I have labeled the orthodox view, is that noth-ing outside the brain is part of it.6

Stated this way, the orthodox view is a metaphysical view about the relationbetween subjective experience and the brain. This metaphysical view is often com-bined with a certain conception of consciousness and how neuroscience shouldinvestigate it. According to this conception, the first main task for the neuroscienceof consciousness is to find the neural correlates of consciousness (the NCC), specifi-cally the minimal neural correlates for the phenomenal contents of consciousness.7

165


A minimal neural correlate for a given conscious experience, such as the visualexperience of the color red, is the minimal set of neuronal events and mechanismsjointly sufficient for that conscious experience.8 According to the NCC researchprogram, the discovery of the NCC should be the first prime goal of the neuro-science of consciousness.

Combining the NCC research program with Brainbound (or what Block callsthe orthodox view) gives us a certain picture of consciousness and the brain. Toexplain this picture, however, we need to introduce some conceptual distinctions.

The first distinction is between state consciousness and creature consciousness;that is, between phenomenally conscious states and phenomenally conscious crea-tures.9 The standard way to explain these notions is to say that a phenomenally con-scious state is a state for which there’s something it’s like for the subject to be in thatstate. Such states are individuated in terms of their phenomenal content or theirphenomenal character. A phenomenally conscious creature is a creature for whichthere’s something it’s like to be that creature.10 Phenomenally conscious creaturesare subjects of experience.

In cognitive and clinical neuroscience, studies of the neural substrates of con-sciousness tend to focus either on state consciousness or creature consciousness.

State-based studies contrast the reportable awareness of a given phenomenalcontent (such as seeing a face) with the lack of reportable awareness of that content(for example, in a masking experiment), or with the reportable awareness of a dif-ferent phenomenal content (such as a house; for example, in a binocular rivalryexperiment). In these studies, subjects are awake, aware, and able to report thechanging contents of their phenomenal consciousness.

Creature-based studies focus either on the contrast between phenomenal con-sciousness and its absence (under anesthesia or during coma) or between differentglobal or background states of consciousness (such as wakefulness and dreaming).

The concepts of creature consciousness and background state consciousness arethus closely related. Background states of consciousness, such as wakefulness anddreaming, are domain general, not modality specific. They characterize one’s over-all phenomenal perspective as a conscious subject. In this way, they are propertiesof conscious creatures.11

Speaking roughly, we can say that state-based studies indicate that thereportable awareness of distinct phenomenal contents depends on the activation ofparticular cortical regions and networks,12 whereas creature-based studies indicatethat subcortical systems such as the thalamus and brain stem are crucial for thepresence of consciousness compared with its absence, and for the transitions anddifferences between wakefulness and other background states such as dreaming.13

The next distinction we need to make is between the core realization and thetotal realization of a given phenomenally conscious state.14 In general, the core real-ization of a property or capacity suffices for that property or capacity only whenproperly placed in the context of a larger system that constitutes the total realiza-tion.15 Block proposes that “the core NCC is the part of the total NCC that distin-

166


guishes one conscious state from another—the rest of the total NCC being consid-ered as the enabling conditions for the conscious experience.”16

According to this proposal, the total NCC comprises the neural substrates ofcreature consciousness; that is, of a subject’s being able to experience any phenom-enal states at all. These substrates can be treated as enabling conditions, however,in relation to the core NCC, which distinguishes one conscious state from anotherin terms of its specific phenomenal content.

Yet this formulation still isn’t complete. In general, the total realization of aproperty or a capacity suffices for that property or capacity only given the appro-priate background conditions.17 Similarly, the total NCC suffices for creature con-sciousness only given certain background conditions, which normally includenonneural parts of the body and the environment.

Let’s connect these points to the metaphysics of Brainbound or the orthodoxview. For Block, the core NCC of a given experience is equivalent to the minimalconstitutive supervenience base for that experience. In other words, fixing the coreNCC fixes the phenomenal content of the experience. As Block says, “if the relevantbrain state were to come about—somehow—the experience would be instanti-ated.”18 In his example, were a disembodied, freestanding brain in the relevant stateto arise through the chance fluctuations of microphysical particles, it would instan-tiate the experience (that is, it would instantiate a conscious state with the samephenomenal content, though not the same intentional or representational content).

The classic philosophical expression of this idea, of course, is the brain in a vat.According to this thought experiment, a disembodied brain placed in a life-sustain-ing vat and stimulated in the right way by a supercomputer would have experienceswith the same kind of phenomenal content or subjective character as you have.

Although usually a philosophical device for raising epistemological questionsabout the relation between subjective experience and the world, the brain in a vatalso functions as a methodological device in philosophical reflection on the neuro-science of consciousness. There it serves to illustrate the idea that the body can be“screened off” from the neural correlates of consciousness, which then get meta-physically conceptualized as the minimal constitutive supervenience base for sub-jective experience. We can see the brain in a vat put to this use in the followingremarks by Thomas Metzinger:

[T]here is a minimally sufficient neural correlate for the content of con-sciousness at any given point in time. If all properties of this local neu-ral correlate are fixed, the properties of subjective experience are fixedas well. Of course, the outside world could at the same time undergoconsiderable changes. For instance, a disembodied but appropriatelystimulated brain in a vat could—phenomenologically—enjoy exactly thesame kind of conscious experience you do right now while reading thisbook.19

What we want to do now is to examine precisely this idea—illustrated force-fully by the brain in a vat—that the body can be screened off from the neural

167


substrates of consciousness. Our next step will be to reconsider the brain-in-a-vatthought experiment. Thinking carefully about this experiment will reveal that thebrain and body are so deeply entangled, structurally and dynamically, that they areexplanatorily inseparable. Dynamic entanglement implies that we can’t understandconsciousness—especially creature consciousness—by considering only the activ-ity of neurons apart from the body, and hence we have good explanatory groundsfor supposing that the minimal realizing system for consciousness includes thebody and not just the brain. In this way, we intend to put the brain-in-a-vatthought experiment to a new use, one that supports the enactive view of experi-ence.20

II. A CLOSE LOOK AT THE BRAIN IN A VAT

What would it take to envat the brain so that it would function exactly like itsembodied counterpart? To our knowledge, the only philosopher who has addressedthis question is Dan Dennett. We’re thinking not of his classic paper “Where AmI?”21 which dramatically portrayed the brain-in-a-vat scenario for cognitive science,but rather of the “Prelude” to his book Consciousness Explained.22 There he addressesthe question from a technical, bioengineering perspective, not the philosophicalperspective of what is conceivable or conceptually possible in principle. We’re goingto follow Dennett’s lead, for as he notes, “sometimes an impossibility in fact is the-oretically more interesting than a possibility in principle.”23 We’ll address the merepossibility in principle of the brain in a vat later.

In thinking about how to envat the brain, we need to consider three mainthings—keeping the brain alive and up and running, the brain’s self-generatedactivity in tight coupling with the body, and what it would take to mimic preciselythe stimulation the nervous system normally receives from the body and the envi-ronment.

2.1 KEEPING THE BRAIN UP AND RUNNING

Before we can appropriately stimulate the envatted brain, we need to keep it aliveand functioning. This already is no mean feat.

First, we need some protective apparatus for the brain. This apparatus servesto replace the skull (and spine, if we choose to keep the spinal cord). To ensure thebrain’s flotation, the protective device will need to be filled with a liquid analogousto the cerebrospinal fluid. This liquid needs to be able to remove waste products ofneuronal metabolism and so must be continually recycled.24 One way to achievethis recycling would be to couple the protective fluid to the second thing we need—a circulatory system.

Almost everyone has experienced the intense dizziness and cognitive impair-ment that happens when you stand up fast. The unimpeded supply of blood to

168


every part of the brain is critical for its functioning and by no means a trivial phys-iological accomplishment. To envat the brain, we must provide an adequate bloodsupply or a fluid with similar biochemical properties. For this task we could prob-ably choose to keep the vascular system in place as a delivery structure. Alterna -tively, in the true spirit of the thought experiment, we can imagine replacing theentire cerebral vasculature with some synthetic device that shows similar proper-ties of selective permeability and local and systemic responsiveness to the brain’songoing demands. This kind of responsiveness is absolutely crucial. Without itthere would be no way to compensate for even minimal departures from home-ostasis due to neuronal activity, with fatal consequences for our experiment.

The tight coupling of blood flow and neuronal activity is a basic physiologicalfact known as functional hyperemia.25 Neuroimaging techniques, such as fMRI andPET, rely on different aspects of this coupling.26 Although the actual mechanismsunderlying the coupling aren’t fully understood, a variety of molecular and cellu-lar factors are known to participate in the regulation of local blood flow in thebrain.

Our life-sustaining system must be able to deal efficiently with these sorts offactors in order to meet the local needs arising from ongoing neuronal activity. Soour synthetic apparatus will probably have to be as sophisticated as an actual vas-cular system in its structural features and functional capacities.

Suppose we’ve succeeded in setting up such an immensely complex system. Wethen need to move the fluid through the delivery structure. Here some kind ofpump is needed, as well as some minimal and highly selective recycling system forreplenishing the fluid’s necessary components, including oxygen, glucose, and thenumerous soluble ions, proteins, and other biomolecules that account for the fluid’sosmotic, nutrient, and regulatory properties.

Because the pump and recycling system must be responsive to the brain’sactual demands, they should be functionally coupled to the brain’s activity. Thiscoupling ensures the local availability of the soluble factors provided by the circu-latory system and keeps the concentration of the circulating molecules and ionswithin a physiological range despite continuous demands from the neuronal tissue.The brain normally relies on a host of regulatory loops involving organs outside thebrain to meet these needs.

Let’s summarize things up to this point. Whatever life-sustaining system weproduce, it will involve at least the capacity to keep up with the brain’s energetic,ionic, osmotic, and recycling needs. It will therefore include some kind of circula-tory system plus the necessary pumps, oxygenating devices, and additional subsys-tems for maintaining physiological levels in the circulating fluid.

The next point is less obvious. What the brain requires at any given instantdepends on its own ongoing and self-generated activity, or what neuroscientists call“intrinsic activity.” Our life-sustaining system must support this intrinsic activityand respond to it locally and systemically at any given instant, independent of anyoutside evaluation of the brain’s needs. Consequently, our life-sustaining system

169


needs the kind of robustness and flexibility we see only in energetically open, self-sustaining, and self-regulating systems—that is, organisms.

This life-sustaining system is starting to look less like a vat and more like a sur-rogate living body or organism.

2.2 SELF-GENERATING ACTIVITY AND THE BODY-COUPLED BRAIN

When considering the requirements of keeping the brain up and running, we beganby taking an external control perspective. From this perspective, the issue is how tocontrol the brain from the outside so that it stays alive and functioning. Yet once wetake into account the brain’s endogenous workings, it becomes clear that our life-sustaining system must be intimately coupled to the nervous system’s own intrin-sic activity.

This basic requirement necessitates a radical shift in how we think about ourvat. Whatever life-sustaining system we construct, the functioning of its every part,as well as its overall coordinated activity, must be kept within a certain range by thenervous system itself in order for the brain to work properly. Hence the externalcontrol perspective is not generally valid. Instead, our life-sustaining system and thebrain must be seen as reciprocally coupled and mutually regulating systems.

According to a number of recent proposals, this kind of tight coupling betweenneural and nonneural factors—between brain and body in the normal embodiedcase—constitutes the organism as a functional unity and underwrites the phenom-enal feeling of self that permeates normal creature consciousness.27 For example,according to Antonio Damasio, the nervous system continually maps the state ofthe body through a series of core neural structures that are crucial for both bodyregulation and the feeling of self.28 In this theory, creature consciousness with aminimal phenomenal feeling of self arises as a feature of life-regulation processeseffected by the nervous system in tight coupling with the body.

Damasio hasn’t missed the implications of his proposal for the brain-in-a-vatthought experiment. In his book Descartes’ Error he writes:

It might be argued that if it were possible to mimic, at the level of thedangling nerves, realistic configurations of inputs as if they were com-ing from the body, then the disembodied brain would have a normalmind. Well, that might be a nice and interesting experiment “to do” andI suspect the brain might indeed have some mind under those condi-tions. But what that more elaborate experiment would have done is cre-ate a body surrogate and thus confirm that “body-type inputs” arerequired for a normally minded brain after all. And what it would beunlikely to do is make the “body inputs” match in realistic fashion thevariety of configurations which body states assume when those statesare triggered by a brain engaged in making evaluations.29

Lawrence Shapiro, in his book The Mind Incarnate, relies on Damasio to defendwhat he calls the “embodied mind thesis,” but he also thinks Damasio’s assessmentof the brain-in-a-vat thought experiment is confused. Damasio seems to be saying

170


that it would be possible to remove the brain without disrupting the type of inputsit receives, but that the brain would no longer function normally. Shapiro asks: “Ifthe inputs to the brain in a vat were exactly what they are to a brain in a body, whatexplains why the envatted brain would have some mind, but not a normal mind?”30

Furthermore, at the end of the passage, Damasio seems to be saying that it wouldn’tbe possible to mimic precisely the inputs that a brain in a body receives, so thethought experiment wouldn’t be possible after all. But as Shapiro says: “Either it ispossible to mimic the inputs to the envatted brain precisely or it is not. If it is, whywouldn’t the brain perform normally? If it is not, then this is just to deny the coher-ence of the thought experiment.”31

There’s a way to understand Damasio, however, that might alleviate some ofthe confusion. The crucial point is that identical inputs don’t suffice for identicalstates. It’s not the case that an embodied brain and its envatted duplicate willremain qualitatively identical simply because they receive identical inputs through-out their lives. The brain isn’t a reflex machine whose activity is externally control-lable through input instructions. Rather, it’s a highly nonlinear and self-organizingdynamical system whose activity exhibits an extreme sensitive dependence on ini-tial conditions. Inputs perturb such complex systems, but don’t specify particularoutcomes. Furthermore, most inputs arise as a consequence of the system’s ownintrinsic activity. Hence to get the body-type inputs to match the normal inputsprecisely would require getting them to match the bodily inputs to the brain thatarise from the brain’s nonlinear and unpredictable intrinsic activity.

So here’s our gloss on Damasio: As a technical matter, it seems highly unlikelywe could achieve such a matching. Indeed, in a universe with stochastic or noisythermodynamic processes, it’s probably impossible, not merely highly improbable.Nevertheless, whether some kind of creature consciousness could be achieved withsomething less than this kind of match—whether it could be achieved throughwhatever body-type inputs we could technically deliver—seems an open question.

With these thoughts we come to our third requirement—mimicking environ-mental stimulation.

2.3 MIMICKING ENVIRONMENTAL STIMULATION

Here our concern isn’t so much creature consciousness but phenomenal state con-sciousness. What would it take to produce conscious states with specific phenom-enal contents like those of normal perception?

The minimal requirement is to deliver stimulation to the neuronal terminalsthat matches sufficiently well the stimulation the brain normally receives from theenvironment.

Let’s not underestimate the complexity of the stimulating devices. Imagine anartificial device capable of stimulating every fiber of the optic nerve in perfect cor-relation with the light pattern of the scene to be recreated, guaranteeing all thedynamic receptive field relations found originally among retinal cells, maintaining

171


perfect synchrony with the brain’s exploratory motor efference signals as its sensorysystems scan through the virtual image, and updating its activity so as to match pre-cisely the sensory reafference (the sensory feedback caused directly by the motorefference). As Dennett notes, this problem will be computationally intractable oneven the fastest computer and storing all the information is ruled out because ofthe combinatorial explosion of possibilities.32

Another crucial requirement is that we can’t allow these stimulating devices todisrupt the life-sustaining system already established. Every stimulation we deliverproduces a departure from homeostasis for which there must be immediate com-pensation or the whole system will crash. So whatever stimulating devices we con-struct, they must be integrated seamlessly into the “vat.”

We now have two basic requirements. On the one hand, the stimulation mustmimic the stimulation the embodied nervous system normally receives. On theother hand, the stimulating devices must not disrupt the life-regulation crucial forthe functional unity of the system and for the stable realization of creature con-sciousness.

These two requirements imply that our artificial stimulating devices need tobe controlled by the brain itself through sensorimotor loops. In other words, weneed to equip our brain in a vat with real (synthetic, not virtual) peripheral senso-rimotor systems.

A significant body of work in computational neuroscience supports this point.This work makes clear that adaptive behavior arises from the dynamical couplingof the nervous system and peripheral sensorimotor systems; it isn’t programmedor commanded by the brain.33 So the sensorimotor interface we give to our envat-ted brain must have sufficient flexibility and processing capability to establish thisdynamical coupling for the brain. The best and probably only way to establish thiscoupling is to give our brain real sensorimotor systems it can control.

Let’s recall that in order to keep the brain alive and functioning, our so-calledvat must already be self-regulating and self-sustaining like a living organism. We’vejust seen that we need to equip this system with synthetic sensorimotor devices thatcan operate with a certain degree of autonomy while tightly coupling with thebrain’s ongoing intrinsic activity. Our so-called brain in a vat now looks like anautonomous sensorimotor agent. In trying to envat the brain, we have wound upwith an embodied agent in the world.

III. THE NULL HYPOTHESIS

On the basis of the foregoing considerations, we can propose the following nullhypothesis for the brain-in-a-vat thought experiment. Any adequately functional“vat” will be a surrogate body. We don’t mean a body like ours in its material com-position, but one sufficiently like ours in its functional organization. At this level,

172


the term “body” means a self-sustaining system (life-regulation) that controls itsown sensorimotor coupling with the outside world (sensorimotor agency). Inshort, the null hypothesis is that the so-called vat would be no vat at all, but ratheran embodied agent in the world.

This line of thought suggests a way to state Enactive as an alternative empiri-cal hypothesis to Brainbound. The enactive hypothesis is that the minimal biologi-cal realizing system for creature consciousness is not the brain (or some neuralsubsystem) but an organism, understood as a self-sustaining system composed of somecrucial set of dynamically entangled neuronal and extraneuronal subsystems. Thishypothesis is the one that needs to be rejected in order to reject Enactive in favor ofBrainbound.

3.1 THE BRAIN IN A VAT AS A SUPERVENIENCE THOUGHT EXPERIMENT

The brain in a vat is an example of what Susan Hurley calls a supervenience thoughtexperiment.34 In this sort of controlled thought experiment, you divide candidateexplanatory factors into an internal set and an external set relative to some bound-ary, such as the skull or skin. You then suppose the internal factors hold constantwhile the external factors vary. In the brain in a vat, neural states are supposed tohold constant across the embodied and envatted brains, while external factors obvi-ously vary.

A crucial supposition of such supervenience thought experiments is that youcan “unplug” internal factors from one set of external factors and “replug” them intoanother. Being unpluggable accordingly provides a critical condition of possibilityfor the thought experiment: If the internal factors can’t be unplugged from theexternal ones, if they vary together across the relevant situations, then they aren’texplanatorily separable and the supervenience thought experiment isn’t possible.

Our examination of the brain in a vat strongly suggests that neural factorsaren’t unpluggable from bodily factors and thus that the two are explanatorilyinseparable. In the range of possible situations relevant to the explanatory frame-work of the neuroscience of consciousness, the brain-in-a-vat thought experiment,strictly speaking, doesn’t seem possible (because the envatted brain turns out to bean embodied brain after all).

We might wonder now whether the thought experiment is even coherent. Ifenvatting the brain in the right way implies embodying the brain, then the thoughtexperiment undermines itself. The whole point of the thought experiment is toremove the world and the body while preserving brain function. Yet to get the brainto be a brain in the right way we have to give it a body and a world.

3.2 CONCEIVABILITY

If you think the important issue is the mere conceivability or conceptual possibil-ity in principle of a brain in a vat, then everything we’ve said so far will probably

173


strike you as irrelevant. If all that matters is conceivability, then we can avail our-selves of whatever conceivable technical resources we need, regardless of whethersuch resources are remotely feasible or even possible in our world or in worlds withour laws of nature.

But such conceivability or possibility in principle tells us virtually nothing ofinterest with regard to what concerns us here, namely, the explanatory frameworkof the neuroscience of consciousness in relation to embodied cognitive science, or to be more specific, Brainbound versus Enactive as rival explanatory research programs.

Consider that even if our concern is mere conceivability, we still have to facethe question of what the term “brain” designates, such that envatting that and onlythat would suffice to duplicate experience. Do we need to envat only the neuronsand their synaptic connections? Or must we also envat the glial cells, which out-number the neurons nine to one and are now believed to be critical to learning andthe formation of memories?35 Do we need to envat immune cells, which togetherwith glial cells form complex cellular and molecular communication networks thatmediate mood, emotion, and pain?36 Are the immune systems and endocrine sys-tems mere boundary conditions on the neuronal determination of consciousexperi ence or do they belong to the minimal realizing system for consciousness? Wesimply don’t know the answers to these questions.

Moreover, if one accepts that there’s an explanatory gap—that we presentlyhave no understanding of how phenomenal consciousness could be explained interms of neural activity—then we can hardly assume that if there were a successfulbiological explanation of consciousness, it would explain consciousness only interms of strictly neural activity instead of appealing to some wider system compris-ing other nonneuronal biological factors.

One might respond to these considerations by saying that, given our presentscientific knowledge, it’s implausible to suppose that the realizing system for con-sciousness includes anything nonneuronal. Jesse Prinz expresses this thought in thefollowing way:

We have never found any cells outside the brain that are candidates ascorrelates for experience. Such cells would have to co-vary with con-scious states in content and time course. Every component of the bodythat we can experience is represented in the brain, and when the corre-sponding brain areas are damaged experience is lost. Conversely, bodilyexperience can continue after the body is damaged, as in the case ofphantom limb pain. There is, in short, little reason to think the corre-lates of experience extend beyond the cranium.37

But this assessment seems to miss the mark. It treats the problem of explaining con-sciousness simply as the problem of explaining phenomenal state consciousness ina given sensory modality and therefore assumes that the issue of embodiment isdecided by the fact that we’ve found cortical neurons whose receptive field proper-ties correspond in certain limited respects with aspects of phenomenal content,38

174


whereas we haven’t found any nonneuronal cells that correlate with conscious statesin this way. Yet in every experiment that establishes these correlations, the subjectsare already conscious and able to report the changing contents of their awareness.Thus, the neural correlates of these phenomenal changes aren’t sufficient for con-sciousness because they presuppose that the subject is already conscious with sometotal field of background awareness.39 The crucial problem is to explain this back-ground consciousness. In other words, the crucial problem is to explain why thecreature is conscious at all. Background states of consciousness, such as waking anddreaming, are global modulations of creature consciousness.40 So the crucial mat-ter is to account for creature consciousness. How can we be so confident that crea-ture consciousness is explainable in strictly neural terms?

According to current evidence, the biological basis for creature consciousnessincludes subcortical parts of the brain that are deeply involved in life-regulationand bodily homeostasis.41 These subcortical structures don’t command the body ina hierarchical, master-slave way. Instead, they modulate bodily processes by beingdensely interconnected to them on multiple cellular and molecular levels. Giventhis dense interconnectivity, it seems that neuronal and nonneuronal factors aren’tunpluggable, and hence aren’t explanatorily separable. As Damasio especially hasemphasized, it makes little biological sense to talk about “what the brain does” ver-sus “what the body does.”42 If the physiological system that supports creature con-sciousness comprises densely coupled neural, endocrine, and immune processes,and if it comprises sensorimotor loops through the body and the environment,then the biological basis of consciousness isn’t brainbound.

The upshot of these considerations for the mere conceivability of a brain in avat is that since we don’t actually know what the minimal biological requirementsare for creature consciousness, we don’t know what we’re supposed to imaginewhen we imagine a brain in a vat, so the mere conceptual possibility of a brain in avat seems an empty scenario.

IV. BRAINBOUND REVISITED

Where do these considerations leave us with regard to Block’s claim that the “mim-imal constitutive supervenience base” for experience is the brain or more preciselythe core NCC?

Here we need to be careful. Neural supervenience for experience requires thatwhen the neural factors are held constant across some range of cases, then so is theexperience. As Hurley observes, this kind of internal supervenience can hold eventhough the corresponding supervenience thought experiment isn’t possible becauseinternal factors and external factors aren’t unpluggable. So the minimal superve-nience of the subjective character of experience on brain states is compatible withthe explanatory inseparability of brain and body states. As Hurley also observes,

175


however, the mere truth of internal supervenience provides no support for inter-nalist explanation, if the relevant supervenience thought experiment isn’t possible,because internalist explanation requires explanatory separability.43

The main moral of our examination of the brain in a vat is that brain states can’tbe unplugged from body states. So internalist explanation isn’t a good framework forthe neuroscience of consciousness. We should prefer Enactive to Brainbound.

But what about Block’s disembodied, freestanding brain that spontaneouslyarises from the chance fluctuations of microphysical particles? We would like toknow more about this brain. Is it just the neurons and synaptic connections, some-how floating independently of its glial cells, cerebral vasculature, immune cells, andother nonneuronal, somatic partners? That seems physically impossible, not merelyhighly unlikely. How would such a system hold together long enough for us to sup-pose any experience could be instantiated?

Even if we were to allow for purposes of philosophical argument the conceiv-ability or bare possibility of some sort of freestanding brain, completely decoupledfrom the body, we see no reason to think there would be any experience present atall. Why should we think that this kind of brain is intelligible as a subject of expe-rience or possessor of creature consciousness? Neuroscience and biology in generalgive us no reason. We see no good scientific motivation whatsoever for the idea. Itstrikes us as simply a holdover of the flight from Cartesian dualism.

Here, then, is the enactive response to Block. If creature consciousness is a life-regulation process of an organism, and if perceptual consciousness is a certain kindof interactive relationship between an organism and its environment, then a dis-embodied brain going through the same sequence of internal states as an embod-ied brain is like a disembodied stomach going through the same sequence ofinternal states as an embodied one. The disembodied stomach isn’t digesting andthe disembodied brain isn’t experiencing, because the necessary contexts of thebody and the environment are missing.

V. THE BANDPASS ARGUMENT

There’s one more argument for Brainbound we would like to consider—AndyClark’s “bandpass argument.” Here’s the argument in his words:

[I]t is plausible that speed (or fine temporal issues more generally)makes a crucial difference in the moment-by-moment construction ofconscious experience itself. Thus suppose conscious experience requirescortical operations that involve extremely precise temporal resolutions,such as the synchronous activations of distinct neural populationswhere the required syncrhony requires millisecond precision . . . . Insuch cases the external environment may well matter insofar as it drivesthe neural systems, but the key effects that enable and explain the qual-

176


ity of felt experience may then be occurring at time-scales that are onlypossible within the neural apparatus itself. If this were so, then every-thing that involves subsequent motor actions or bodily action (includ-ing active saccades around the scene) will be “screened off ” (by thebodily “low-pass filter”) from the neural/CNS mechanisms that actuallyproduce the conscious experience.44

The argument is that the extra-neural body acts as a low-pass filter for signals com-ing from the environment—in other words, the body admits only slow frequencysignals to arrive at the brain—but the contents of conscious experience require fasttime-scales on the order of milliseconds, and the only locus where such fast fre-quency processing can occur is inside the brain. So brain processes directly deter-mine conscious experience, and the body contributes only causally but notconstitutively to consciousness.

We see several problems with this argument.First, there’s a basic empirical problem. The time it takes for visual stimulation

to pass through the lens of the moving eye and reach the first stages of visual pro-cessing is a fraction of the time it takes for neural systems to build up any correlatedactivity (from retina to early visual areas, and then to recurrent loops with highervisual areas and long-range coordination with parietal and frontal regions). Sotreating the body as a low-pass filter in relation to the brain doesn’t work.

Second, Clark seems to be working with a conception of perceptual experiencethat enactive theorists reject. He seems to conceive of a perceptual experience as aninternal state—or the content of a perceptual state as the content of an internal rep-resentation—whereas enactive theorists argue that to perceive is to be in an inter-active relationship with the world. From the enactive perspective, all the bandpassargument shows is that fast frequency brain processes are a necessary part of thebiological basis of the interactive relationship that constitutes perception, but notthat these processes are minimally sufficient for conscious perception (or directlydetermine or constitute conscious perception). Indeed, it’s hard to see how theargument could show that fast frequency synchronous oscillations directly deter-mine perception unless one has already assumed that perception is an internalepisode rather than an interactive relationship with the world.45

Finally, Clark focuses entirely on phenomenal state consciousness and neglectscreature consciousness. Although modality-specific perceptual contents come andgo on a fast time-scale, they do so against the much more slowly changing anddomain-general background state of waking creature consciousness. The crucialconcern of the enactive view is to account for creature consciousness. Clark, how-ever, says nothing about creature consciousness and seems not to recognize that it’sthe crucial issue. In any case, the bandpass argument says nothing against the enac-tive view that the body belongs to the minimal biological requirements for creatureconsciousness.

177


VI. CONCLUSION

The main conclusion we would like to draw from this paper is that we shouldn’tsaddle the neuroscience of consciousness with Brainbound or what Block presentsas the metaphysics of the orthodox view. Research on the neural correlates of con-sciousness provides a wealth of interesting and important findings about the brain.It’s a mistake, however, to think that this research amounts to a search for a mini-mal constitutive supervenience base for experience in the brain. On the contrary,this kind of heavy philosophical interpretation goes way beyond anything neuro-science gives us reason to believe. More important, it also hinders research becauseit prematurely shuts down a whole range of important and wide-open questionsabout brain-body relations in the biological realization of consciousness.

One of these questions—the one that’s concerned us here—is whether thebrain alone suffices for creature consciousness or whether the body is also required.We’ve given reasons to think that the body and brain are so dynamically entangledin the causation and realization of consciousness as to be explanatorily inseparable.If these reasons are sound, then we should prefer Enactive to Brainbound. In otherwords, we shouldn’t expect there to be a purely neural explanation of conscious-ness. Instead, what we should expect—or at least aim for—is a much richer biolog-ical account of consciousness as a life-regulation process of the whole organismdynamically engaged in its world.

ACKNOWLEDGMENTS

For helpful comments and criticism, we are indebted to the Philosophy andNeuroscience Reading Group at the University of Toronto, especially Stephen Biggs,Nigel Francis, Matthew Fulkerson, Mike Lachelt, and Adrienne Prettyman; thanksalso to Tim Bayne, Dorothée Legrand, Thomas Metzinger, and Mog Stapleton.

NOTES

1. See Evan Thompson, Mind in Life: Biology, Phenomenology and the Sciences of Mind (Cambridge,MA: Harvard University Press, 2007). See also Diego Cosmelli and Evan Thompson, “Embodi -ment or Envatment? Reflections on the Bodily Basis of Consciousness,” in Enaction: Toward a NewParadigm for Cognitive Science, ed. John Stewart, Olivier Gapenne, and Ezequiel Di Paolo(Cambridge, MA: MIT Press, 2010), 361–85.

2. We borrow the term “brainbound” from Andy Clark, Supersizing the Mind: Embodiment, Action,and Cognitive Extension (Oxford: Oxford University Press, 2008). Clark uses it as a name for thethesis that the material vehicles of cognitive processes reside exclusively inside the skull.

3. Ned Block, “Review of Alva Noë, Action in Perception,” Journal of Philosophy 102 (2005): 259–72.

4. Alva Noë, Action in Perception (Cambridge, MA: MIT Press, 2004).

5. Ibid., 25.

178


6. Block, “Review of Alva Noë, Action in Perception,” 264.

7. See David J. Chalmers, “What Is a Neural Correlate of Consciousness?” in Neural Correlates ofConsciousness, ed. Thomas Metzinger (Cambridge, MA: MIT Press, 2000), 18–39.

8. Christof Koch, The Quest for Consciousness: A Neurobiological Approach (Englewood, CO: Robertsand Company, 2004), 16.

9. See Tim Bayne, “Conscious States and Conscious Creatures: Explanation in the Scientific Studyof Consciousness,” Philosophical Perspectives 21 (2007): 1–22.

10. Here we follow Tim Bayne in not defining creature consciousness as wakefulness, but rather asbeing a phenomenally conscious creature. On the one hand, phenomenal consciousness can occurduring sleep, notably while dreaming. On the other hand, patients in a vegetative state can beawake but apparently lack phenomenal awareness.

11. See Bayne, “Conscious States and Conscious Creatures: Explanation in the Scientific Study ofConsciousness,” 13–14.

12. See G. Rees, G. Krieman, and C. Koch, “Neural Correlates of Consciousness in Humans,” NatureReviews Neuroscience 3 (2002): 261–70.

13. See J. Parvizi and A. Damasio, “Consciousness and the Brainstem,” Cognition 79 (2001): 135–59.

14. See Chalmers, “What Is a Neural Correlate of Consciousness?” and Ned Block, “Two NeuralCorrelates of Consciousness,” Trends in Cognitive Sciences 9 (2005): 46–52.

15. See Sidney Shoemaker, “Some Varieties of Functionalism,” Philosophical Topics 12 (1981): 93–119;Robert A. Wilson, “Two Views of Realization,” Philosophical Studies 104 (2001): 1–30.

16. Block, “Review of Alva Noë, Action in Perception,” 47.

17. See Robert A. Wilson, “Two Views of Realization.”

18. Ned Block, “Review of Alva Noë, Action in Perception,” 265.

19. Thomas Metzinger, Being No One: The Self-Model Theory of Subjectivity (Cambridge, MA: MITPress, 2003), 547.

20. Here we draw from Cosmelli and Thompson, “Embodiment or Envatment? Reflections on theBodily Basis of Consciousness.”

21. Daniel C. Dennett, “Where Am I?” in Brainstorms: Philosophical Essays on Mind and Psychology(Cambridge, MA: MIT Press/A Bradford Book, 1978), 310–22.

22. Daniel C. Dennett, Consciousness Explained (Boston: Little Brown, 1991).

23. Ibid., 4.

24. See P. D. Brown, S. L. Davies, T. Speake, and I. D. Millar, “Molecular Mechanisms of CerebrospinalFluid Production,” Neuroscience 129 (2004): 957–70; H. Davson and M. B. Segal, “Secretion andDrainage of the Cerebrospinal Fluid,” Acta Neurologica Latinoamericana 1 (1971): Suppl 1:99–118; M. B. Segal, “Extracellular and Cerebrospinal Fluids,” Journal of Inherited MetabolicDisease 16 (1993): 617–38.

25. See F. Hyder, R. G. Shulman, and D. L. Rothman, “Regulation of Cerebral Oxygen Delivery,”Advances in Experimental Medicine and Biology 471 (1999): 99–110; M. E. Raichle and H. L. Stone,“Cerebral Blood Flow Autoregulation and Graded Hypercapnia,” European Neurology 6 (1971):1–5; R. G. Shulman, F. Hyder, and D. L. Rothman, “Biophysical Basis of Brain Activity:Implications for Neuroimaging,” Quarterly Review of Biophysics 35 (2002): 287–325.

26. See N. K. Logothetis and J. Pfeuffer, “On the Nature of the BOLD fMRI Contrast Mechanism,”Magnetic Resonance Imaging 22 (2004): 1517–31.

27. See Parvizi and Damasio, “Consciousness and the Brainstem”; A. D. Craig, “How Do You Feel?Interoception: The Sense of the Physiological Condition of the Body,” Nature ReviewsNeuroscience 3 (2002): 655–66; Antonio R. Damasio, The Feeling of What Happens: Body andEmotion in the Making of Consciousness (New York: Harcourt, 1999); J. Panksepp, “ThePericonscious Substrates of Consciousness: Affective States and the Evolutionary Origins of Self,”Journal of Consciousness Studies 5 (1998): 566–82; C. B. Saper, “The Central Autonomic NervousSystem: Conscious Visceral Perception and Autonomic Pattern Generation,” Annual Review ofNeuroscience 25 (2002): 433–69.

28. Parvizi and Damasio, “Consciousness and the Brainstem.”

179


29. Antonio Damasio, Descartes’ Error: Emotion, Reason, and the Human Brain (New York: HarperPerennial, 1995), 228.

30. Lawrence A. Shapiro, The Mind Incarnate (Cambridge, MA: MIT Press, 2004), 218.

31. Ibid.

32. Dennett, Consciousness Explained, 5.

33. See H. Chiel and R. Beer, “The Brain Has a Body: Adaptive Behavior Emerges from Interactionsof Nervous System, Body and Environment,” Trends in Neurosciences 20 (1997): 553–57; V. Brezina, C. C. Horn, and K. R. Weiss, “Modeling Neuromuscular Modulation in Aplysia. III.Interaction of Central Motor commands and Peripheral Modulatory State for Optimal Behavior,”Journal of Neurophysiology 93 (2005): 1523–56.

34. Susan Hurley, “The Varieties of Externalism,” in The Extended Mind, ed. Richard Menary(Cambridge, MA: MIT Press, 2010), 101–54.

35. See R. D. Fields, “The Other Half of the Brain,” Scientific American 290, no. 4 (April 2004): 54–61.

36. See S. F. Maie and L. R. Watkins, “Cytokines for Psychologists: Implications of Bi-DirectionalImmune-to-Brain Communication for Understanding Behavior, Mood, and Cognition.Psychological Review 105 (1998): 83–107; J. Scholz and C. J. Woolf, “The Neuropathic Pain Triad:Neurons, Immune Cells, and Glia,” Nature Neuroscience 10 ( 2007): 1361–68.

37. Jesse Prinz, “Is Consciousness Embodied?” in The Cambridge Handbook of Situated Cognition, ed.Philip Robbins and Murat Ayde (Cambridge: Cambridge University Press, 2009), 419–36.

38. For critical discussion of these findings, see Alva Noë and Evan Thompson, “Are There NeuralCorrelates of Consciousness?” Journal of Consciousness Studies 11 (2004): 3–28, and “Sorting outthe Neural Basis of Consciousness. Authors’ Reply to Commentators,” Journal of ConsciousnessStudies 11 (2004): 87–98.

39. John R. Searle, “Consciousness,” Annual Review of Neuroscience 23 (2000): 557–78.

40. See Bayne, “Conscious States and Conscious Creatures: Explanation in the Scientific Study ofConsciousness.”

41. See Parvizi and Damasio, “Consciousness and the Brainstem.”

42. Damasio, The Feeling of What Happens. See also Shapiro, The Mind Incarnate, 217.

43. Hurley, “The Varieties of Externalism,” 7–8. Hurley and Noë deny internal supervenience. Byappealing to cases of neural plasticity, they argue that experiences with the same phenomenalquality can have variable neural correlates. Fixing the neural states does not fix the experience, forthe phenomenal quality depends on the dynamic sensorimotor pattern in which the neural statesparticipate. For example, the neural correlates of seeing an object as white differ before and afteradaptation to colored goggles that distort the wavelengths reaching the eye as a function of eyemovement or object movement across the midline. Although we agree that dynamic sensori motoraccounts of perception provide better explanations of such phenomena than purely neuralaccounts, we are not persuaded that such phenomena count against internal supervenience assuch. What look like variable neural correlates could still share some higher-order neural invari-ance, which might then qualify as a supervenience base for the experience. We cannot see that thearguments from plasticity as developed so far rule out this possibility.

44. Andy Clark, “Spreading the Joy: Why the Machinery of Consciousness Is (Probably) Still in theHead,” Mind 118 (2009): 963–93, at 984–85.

45. For a different view of the role of synchronous neural oscillations in perception, see EvanThompson and Franscisco J. Varela, “Radical Embodiment: Neural Dynamics and Consciousness,”Trends in Cognitive Sciences 5 (2001): 418–25.

180


thompson - brain in a vat or body in a world

Documents